glangdale

127 ms·

1111.
by glangdale 10y ago
I had a very different personal experience, but Karpathy did a wonderful PhD and I did a very marginal one (at a good school, which in many ways makes it worse). From the perspective of the "anchor" rather than the valedictorian,
1112.
by glangdale 10y ago
CUDA is pretty neat; I worked with an early beta (0.7?) and have tracked it on and off over time. It's not magic, though. The point is that GPGPU is really great for some things and pretty ordinary for others. 'Embarrassingly para
1113.
by glangdale 10y ago
We have certainly seen a number of cases where the world's most naive scalar algorithm is compared with a sophisticated GPU-based algorithm to (nominally) do the same thing. It's an easy trap to fall in to - no-one gets a pat on t
1114.
by glangdale 10y ago
We (the Hyperscan team) have spent a lot of time staring at regular expressions over the years (shameless plug: https://github.com/01org/hyperscan ). I think a better format for regex is long overdue, but this isn'
1115.
by glangdale 10y ago
This was my conscious weight loss strategy. Stop eating at 7-8pm in the evening and hold off eating any food until noon or so, which gets a 16+ hour fast. I ate whatever I wanted (not crazy, but not spartan) and lost about 70g per day ove
1116.
by glangdale 11y ago
As for our secrets - they are not really secrets; anyone who wants to learn them can simply read our code. I will outline the literal discovery approach, though, as I'm inordinately proud of it (possibly for the wrong reasons): We find
1117.
by glangdale 11y ago
Your observation about match counts and "all-matches" vs "leftmost-longest" or "pcre-locally-maximizing" semantics is accurate and Hyperscan is meant to be that way. All-matches is the only semantics we can imp
1118.
by glangdale 11y ago
Thanks for looking into this; we like to see benchmarks run by other people to avoid the prospect that we're just benchmarking the things we already know about and like doing (which are often also the things which we've optimized)
1119.
by glangdale 11y ago
Thanks for that story - I'll look up that Amazon link. I love the movie but definitely prefer the book. I'm not sure Red is a better character inherently but I'd certainly rather hang out with him than "Stalker". Ta
1120.
by glangdale 11y ago
We're interested in how the Hyperscan library (full disclosure: I am an Intel employee and Hyperscan designer) stacks up in this comparison. From what I understand there are Go and Rust bindings for Hyperscan out there as 3rd party pro
1121.
by glangdale 11y ago
I enjoyed Stalker, Zona (and those games), but the book is by far my favorite of any of them. The book centers around someone the Strugatsky brothers (authors both of the book and the screenplay, I believe) called "Tough Guy Stalker&qu
1122.
by glangdale 11y ago
It's difficult to separate out the all-important "I did/didn't spend a bunch of time in my teens programming" factor, and if she is getting some attention to fix the temporary deficit that this causes, so much the
1123.
by glangdale 11y ago
The article generally appears to be written with the assumption that the readers already are quite aware of the doings of the Cambridge Five - a reasonable assumption for people over a certain age, but probably less and less likely.
1124.
by glangdale 11y ago
$100M now? Some people are apparently harder to please than the sage and former governor of California: "Money doesn't make you happy. I now have $50 million but I was just as happy when I had $48 million." - Arnold Schwarzen
1125.
by glangdale 11y ago
This is true. One question I've always had about RE2's set interface: if I remember correctly, it reports only true/false for whether the input matches each regex but not number of matches or start and/or end offsets of
1126.
by glangdale 11y ago
I think we'll need to take this discussion offline, and we will be posting more details about how Hyperscan is implemented to our blog over the next few weeks. Our different implementations clearly start with different emphasis; some
1127.
by glangdale 11y ago
We really like to stick to the original meaning of NFA, so our NFAs have the same problems as a DFA from a SoM perspective. That is, our NFAs are just bitvectors over states, not threads that can hold their own SoM or information about pr
1128.
by glangdale 11y ago
Sorry - might have sounded a bit defensive there. We've seen a few systems that have a tendency to sweep correctness under the carpet with shortcuts, especially when you hit nasties like "large bounded repeats in streaming mode&qu
1129.
by glangdale 11y ago
We sacrifice features and compatibility. We have sacrificed zero correctness as far as we are aware. We maintain wart-for-wart compatibility with libpcre in order to have a good basis for comparison for fuzzing and we believe we 100% recogn
1130.
by glangdale 11y ago
Hyperscan is likely a bit heavyweight for this application. Anything that only needs to keep up with user input for a small number of patterns would probably be better off with a lightweight solution with wider pattern support. Although we
1131.
by glangdale 11y ago
Hi - Hyperscan lead architect here. Performance numbers will follow but we do need to be careful about sweeping claims. With the variety of regular expression workloads making a blanket claim about performance is like saying "my C comp